Tesla (TSLA.US) Q3 Conference Call: FSD Supports Demand, Robotaxi and Optimus Still Racing Against the Timeline

According to Financial Associated Press, Tesla (TSLA.US) has released its Q3 financial results and held a conference call. The report shows that Tesla's Q3 revenue was $28.05 billion, flat year-on-year and slightly below market expectations of $28.13 billion. Adjusted EPS came in at $0.48, beating market expectations of $0.44; adjusted EBITDA was $3.9 billion, higher than the market's $3.66 billion forecast. Free cash flow reached $1.53 billion, better than anticipated.

Management described this quarter as characterized by "strong demand and accelerated investment," but the core message is an overall delay in the timeline: FSD is identified as the current largest driver of demand, with over 75% installation rate in North America deliveries and more than 1.7 million paying customers globally. Robotaxi is rapidly expanding in Austin and will enter Dallas next month, but material financial contributions are expected only after Q2 2027. The Optimus production line has entered early-stage production, with deliveries to external customers expected by the end of next year. Achieving a rate of 20,000 units per week (annualized one million units) will take at least another year. Terafab’s equipment cost is several tens of billions of dollars, with the long-term total cost exceeding $100 billion; the first test chip is scheduled for May or June next year, with mass production by the end of 2028. Annual capex guidance has been revised down to around $20 billion, which the company says reflects "efficient capital deployment management" and will continue to expand long-term. The company has secured a credit line of up to $30 billion; no further comments were made on potential mergers.

CEO Elon Musk, CFO Vaibhav Taneja, VP of Autopilot Ashok Elluswamy, VP of Vehicle Engineering Lars Moravy, Konstantinos Laskaris, and Head of Investor Relations Travis Axelrod attended the call. The following are the details of the meeting.

I. Opening Remarks by CEO Elon Musk

Q3 performance was robust, with demand surpassing production. Musk stated this was a "very strong" quarter, with demand growing rapidly and significantly outpacing production. The company is substantially ramping up capacity. FSD is the largest demand driver—"Among all cars available for purchase, FSD has no real competitor." The company is pushing for FSD approval in more markets—several European countries have granted approval, and Musk hopes for EU-wide approval within months, saying "I’d be surprised if it is delayed past Q1."

Semi and Roadster. Q3 saw the start of Tesla Semi truck production and a revamped Roadster demo, which Musk called "one of the coolest product demos ever," a result of collaboration between Tesla and SpaceX teams.

Production Capacity Target: About 5 million units/year by 2030. Musk said the company is laying the groundwork for increased manufacturing capacity and believes there is a real opportunity to achieve this target.

Energy: Megapack Almost Sold Out for 2027. Energy demand is extremely strong—"Our Megapacks for 2027 are almost completely sold out." Production is being ramped up as quickly as possible, and gross margins are expected to rise significantly.

Robotaxi: Unsupervised Mileage Growing Exponentially. Musk said the total fully autonomous driving miles driven by Robotaxi are scaling "exponentially," with weekly growth that is "difficult to overstate." "No other company in the world can match our scale in autonomous driving." Since the Austin Cybercab fleet went live about a month ago, it has grown nearly tenfold. The weekly cumulative mileage chart "looks like a rocket launch."

Optimus: Early Production, S-curve in Initial Slow Stage. The high-volume Fremont production line has begun "early production." Musk emphasized that Optimus’s growth follows an S-curve, but the initial stage has a gentle, bumpy slope because "10,000 things must be done right simultaneously, and production speed depends on the slowest and most challenging of those." The bulk of the "1 million units/year" line is already set up at Fremont, but achieving an annualized output of 1 million units or 20,000 units per week "will take at least a year, maybe longer," as design is still evolving. Musk expects AI, electromechanical design, reliability, and lifespan issues to be "resolved around the end of next year," at which point deliveries of Optimus to non-Tesla entities will begin. Initial customers will be in industrial applications—dangerous, repetitive, undesirable, and 24/7 required tasks.

Terafab: Equipment Commissioning in H1 Next Year, First Test Chip in May/June. The Austin R&D semiconductor fab is progressing rapidly, with equipment commissioning starting in H1 next year and the first test chip expected in May or June—"just to see if we can make a chip at all." The goal is some form of meaningful chip mass production by around Q3 2028. Musk said Terafab "will be the most valuable building in the world," with economic value ranked in the top 20 countries globally. It is the key bottleneck solution for scaling up businesses like Optimus (memory, packaging, logic chips, etc.).

Mergers: “As for the potential merger with SpaceX, I said everything I needed to say yesterday—dancing in a minefield. No more comments.”

II. Remarks by CFO Vaibhav Taneja

Deliveries: Two Consecutive Quarters Above 480,000 Units; Highest Backlog in Four Years. For the first time, deliveries topped 480,000 units in back-to-back quarters. End-of-quarter backlog is at a four-year high—deliveries are mainly limited by production capacity instead of demand. The company is focused on ramping output at all factories.

Model Y Outstanding. Preliminary data shows Model Y is the best-selling BEV in Europe and the top-selling passenger car in South Korea, setting delivery records in markets like Australia, Japan, and Taiwan.

FSD: Over 75% Take Rate in North America, 1.7+ Million Paying Customers Globally. In Q3, over 75% of North American deliveries had a paid FSD subscription at delivery—about 20 percentage points higher than Q2. Worldwide, paying FSD customers exceeded 1.7 million, half of whom are now subscribers for the first time. As FSD performs strongly in the US, the company is taking extra steps in Europe, offering promotions on some models to expand the user base before full FSD approval.

Automotive Gross Margin (Ex Credits): 16.3%→18.1%. The q/q increase was mainly driven by a higher proportion of APAC deliveries. Commodity price increases and rising interest rates continue to be headwinds, partially offset by adjustments to vehicle financing options.

Energy: Deployed 13.7 GWh, Gross Margin 20.4%→33.1%. This was the second-best quarter for deployment in history. Gross margin improvement q/q was mainly due to tariff refunds recognized in the quarter and a warranty reversal after reaching a compensation agreement with suppliers on some cell costs. Data center demand is extremely strong; the company is doing its best to expand capacity, and expects ASP and gross margin to continue to improve.

Services & Other: Gross Margin 14.1%→15.9%, Record High. Driven by multiple businesses including Supercharging and insurance. Robotaxi (grouped under Services & Other) continued to grow with the Austin Cybercab launch; unsupervised mileage grew 8x during the quarter, with double-digit weekly growth.

Expenses & Net Income: The company is in a heavy investment cycle with opex continuing to rise q/q, mainly from SI R&D (depreciation of new compute, chip development) and ramp costs for new products like Optimus, Megapack 3, and Semi. This trend will last until these products are commercialized. Net income was affected by a $142 million reduction due to mark-to-market accounting for SpaceX holdings, which was offset by approximately $290 million of mark-to-market gains on bitcoin.

Capital Expenditure: Full-year Guidance Lowered to Around $20 Billion. This is due to "efficient capital deployment management," and will mainly rely on operating cash flow. Free cash flow exceeded $1.5 billion at the end of the quarter, with cash & investments above $45 billion. A credit line up to $30 billion has also been secured to support anticipated capex growth; no new drawdowns are planned this year. Tariff refund proceeds are being reinvested into expanding US production capacity. In the coming years, capex will continue to grow to expand Robotaxi fleets, increase Optimus capacity, install solar manufacturing capacity, purchase SI compute infrastructure and build semiconductor fabs.

III. Q&A Session

1. How will Cybercab expand beyond Austin? Regulatory pathway under NHTSA review?

Ashok Elluswamy: Cybercab was launched this quarter and mainly learns from data from other platforms like Model Y/3, so it does not require a large number of vehicles to acquire data. Austin is only the starting point—"There is no fundamental reason to limit ourselves to Austin." Dallas will launch next month; the same software can scale to many other cities and be used with Model Y Robotaxi as well.

2. For Optimus, is training data the main bottleneck? Crowdsourced data vs internal Optimus Academy?

Musk: Crowdsourcing is not the problem. There's still a lot of AI architecture work to do. For Optimus to get training data, about 10,000–20,000 robots need to be manufactured to engage in self-play and build the Optimus Academy to practice various tasks and form a data flywheel. Unlike cars, Optimus doesn’t have a massive fleet logging data, so it must "bootstrap" via manufacturing robots.

Ashok adds: Cars have native data; Optimus doesn’t have an exoskeleton people can "wear" as a puppet. Optimus must first learn from observing humans, then practice on its own. The company is setting up large-scale reinforcement learning: basic training from human data, then massive self-play in both simulation and the real world.

Musk: We may release an Optimus video in a month or two showing non-preprogrammed tasks, not just "canned demos." A key milestone is integrating the AI5 chip into Optimus around April next year, with an order-of-magnitude performance boost over AI4, which will hugely help since humanoids have far more degrees of freedom than cars.

3. When will consumer vehicles have access to unsupervised FSD?

Ashok: Cybercab and Model Y use the same software; version 15 will be released to customers. An additional requirement is that customer vehicles connect to Tesla's geofencing and fleet management systems. The first customer vehicles may join the Robotaxi fleet around Q2 next year.

4. When will Cybercab be sold to consumers?

Lars Moravy: The plan is to deliver Cybercab to customers before the end of 2026, but currently all Texas-built vehicles go to Tesla’s own Robotaxi fleet. As Tesla grows, most fully designed autonomous vehicles delivered to customers will be Cybercabs. We’ll start small this year, learning about delivery, ownership, and fleet management, then scale up gradually.

Musk adds: Cybercab is mainly a "transportation-as-a-service" offering; the vast majority of Cybercabs will be managed directly by Tesla as Transport-as-a-Service. People who want to own a car will be more interested in Model 3/Y and similar vehicles.

5. Robotaxi unit economics: cost per paid mile & Optimus unit cost?

Vaibhav: Utilization in core Robotaxi operating areas has improved; weekly paid orders and mileage are growing. We haven’t reached the $0.30/mile target yet but see a clear path, with scale being the key. Cybercab is only in Austin now, but will expand to other cities by year-end. As the fleet grows, we are learning about fleet management, pickups/drop-offs, cleaning, operations, and optimizing costs.

Optimus cost: Internal tracking of design maturity is progressing well, but it’s too early to say if we’re at the cost target since designs are still iterating. Cost-wise, we’re trending toward final Cybercab cost levels.

6. FSD map issues: Why not use your own maps?

Ashok: Agreed, reliance on external maps should be avoided. We’re building our own map—not in the style of OpenStreetMap, but a new AI-powered map and navigation architecture that integrates voice, personal preference learning, and fleet-wide intelligence. All vehicles’ collective intelligence will feed our maps and navigation. Version 15 will first use it for trip start/end, with full navigation, parking, and destination management rolling out in the coming months.

7. Existing vs planned factory capacity?

Lars: Factories ran at record speeds this quarter, but capacity is limited by the slowest process or component. Over the past year, we’ve made strategic global investments to boost capacity, whose effects are showing this quarter and should continue over the coming quarters. We’re ramping Cybercab at Giga Texas, have applied for a Berlin factory expansion permit, and Reno’s new Semi truck factory has launched. We are definitely building new factories, aiming for 5 million units by 2030.

8. FSD competition and pricing: Will other automakers’ L2+ products create price pressure?

Ashok: The first 90% of cool demos is easy—solving long tail issues is extremely hard, especially reaching safe, injury-free millions of miles. As unsupervised FSD rolls out to the wider customer fleet, "I don’t think anyone will match us in the coming years."

Vaibhav: In US showrooms, customers now ask "can I try FSD?" rather than asking about Model 3/Y or Cybertruck. FSD is now the key driver, giving back travel freedom to people who previously couldn’t drive.

About China: Compared with Chinese OEMs, Tesla still maintains a lead in technology today. The company is working hard for broader FSD approval in China; some models are already FSD-enabled. Competing CEOs who test FSD in the US are reportedly shocked. The company is actively cooperating with Chinese regulators and has published FSD safety and unsupervised Robotaxi safety reports.

9. Robotaxi revenue & profit timeline? Cybercab line ramp curve?

Vaibhav: By exponential math, Robotaxi will become a significant revenue and profit contributor in 2027.

Musk: "It won’t change the game this quarter. We’ll probably see an obvious impact starting Q2, possibly becoming a major event in Q3 or Q4."

On Cybercab line: The engineering team designed the product for highly scalable manufacturing; capacity is not the challenge—the key is regulatory approval for immediate deployment after production. Cybercab has 50% fewer parts than any other model, and actual line capacity far exceeds the listed 125,000 units/year—currently producing to match Robotaxi service demand.

10. Energy business: Role of distributed generation & distributed compute? Megapack?

Musk: It’s still early, and the full picture is unclear. By Q1 or Q2 next year, things should be clearer. The potential is huge, but it’s hard to say how big right now.

11. Is the bottleneck for Robotaxi growth regulatory or software-related?

Ashok: There are still software challenges. Robotaxi service hours have extended from 6 am to 9 pm, moving to 10 pm, soon 2 am—gradually increasing availability, driven by AI software capability. Zero safety incidents is the guiding principle; even the slightest risk will not be accepted. This year will unlock 24/7 operation and higher-speed roads, which are necessary for scaling. We’re expanding both breadth (6 cities) and depth (dense fleet in Austin), with plans to cover more weather, day/night conditions, highways, and high-speed scenarios by the end of this year. After that, it’ll be more about regulatory rather than software hurdles—Q1 or Q2 we expect the shift.

12. How much capital does Terafab need? Milestones?

Musk: Capital requirement is very large—equipment cost will be tens of billions of dollars, and the cost split between Tesla and SpaceX is still being determined. By early 2028, equipment capex won’t be less than several tens of billions; the long-term total cost may exceed $100 billion.

Vaibhav: This is Tesla’s first semiconductor fab—the majority of equipment orders for the first set have been placed. Mid-next year is a key development milestone. Real mass production won’t happen until late 2028, which is "lightning fast" by fab standards.

13. Biggest design issue with Optimus?

Konstantinos Laskaris: The hand is one of the most challenging hardware areas. Most of the electromechanical system is fairly reliable, but the hand is more complex than the actuators for locomotion, with some specialized gears still being developed for durability and reliability.

Musk: The main challenges are the hand and fingers’ fine design. Humans self-heal but robots do not; human fingers are soft with many sensors inside the skin—it’s extremely hard to build hands that are both abrasion-resistant, capable of delicate grasping, and don’t damage sensors. Also, integrating the AI5 chip is another huge step, with a 10x boost in compute power. The combination of electromechanical dexterity, AI software, and AI chip will see Optimus’s value grow exponentially—"At some point in future, the number of humanoid robots on Earth will exceed that of human beings."

14. Is Robotaxi’s lower pricing than competitors a long-term strategy?

Vaibhav: Tesla lacks drivers and builds its own electric cars, so its cost base is inherently different from ride-hailing companies with drivers. The goal is to "take over all transportation."

Musk: Unless prohibited by government action (which could happen to protect domestic industries), Tesla will cover most passenger transport mileage in any country that allows free operation. "Mark my words on this—I am confident."

15. Can unsupervised mileage maintain double-digit weekly growth? What are the growth drivers?

Ashok: Expansion will continue for the foreseeable future, with no obstacles in sight—a clear path to scaling to millions of cars in the coming months. This is the fruit of years of technical investment and the V15 software stack, with foundations in end-to-end neural networks and human learning. The same tech should scale to millions of vehicles wherever operation is allowed.

16. Is FSD ARR close to $1 billion? Renewal rate?

Vaibhav: Analyst estimates (about $1 billion ARR) are correct. As scale grows, we will begin to disclose more FSD metrics beyond revenue, including other key indicators. We’re not disclosing churn yet due to active free trials in many new countries. The 75% North America take rate includes a free one-month FSD subscription with every new car, yet people keep subscribing.

17. Can Optimus scale production for training even before hand design is finalized?

Musk: That’s exactly what we’re doing—making robots without perfect hands, while simultaneously iterating on Optimus training, supply chain, and production systems. Fremont production lines, supply chain, design engineering, and AI software are advancing simultaneously—"growing as fast as humanly possible." Expect initial customer deliveries around the end of next year. "If anyone can build a robot competitive with Optimus, I’ll be shocked."

Vaibhav: Besides Fremont, a new Optimus factory is being built in Austin with much higher capacity.

18. Semiconductor equipment capacity & capex?

Musk: Not a significant impact at present. Capex is actually slightly below expectations, but will become more notable next year—and reach tens of billions by 2028.